Research

CISA Says REDCap CVE-2026-90817 Is Exploited. Upgrade to 16.0.49 LTS, 17.3.10 LTS or 17.4.4 by Saturday, Nov 21. The Only Key It Needs Is a Survey Link You Already Published.

CISA Says REDCap CVE-2026-90817 Is Exploited. Upgrade to 16.0.49 LTS, 17.3.10 LTS or 17.4.4 by Saturday, Nov 21. The Only Key It Needs Is a Survey Link You Already Published.

If you run REDCap, upgrade today to 16.0.49 LTS, 17.3.10 LTS or 17.4.4 Standard Release, whichever matches the track you are on. Those are the fixed versions named in the CVE record that Securifera, the assigning authority, published on September 20, 2026: "Upgrade to REDCap 16.0.49 LTS, 17.3.10 LTS, or 17.4.4 Standard Release, as applicable."

CISA added CVE-2026-90817 to its Known Exploited Vulnerabilities catalog today, Wednesday, November 18, 2026. The due date for federal civilian agencies is Saturday, November 21, 2026. Three days, and the third one is a weekend. CISA also marked this entry for forensic triage under BOD 26-04, which means agencies are expected to look for compromise, not only patch.

If you cannot upgrade today, the researcher who found the flaw recommends temporary containment: restrict or disable public surveys, and block survey-context access to the root dispatcher, the Data Import controller and the Parcel controller. Treat that as a bridge to the upgrade, not a substitute.

Why "requires a valid public survey hash" is not a lock

CISA's entry ends with the sentence that will make some administrators relax: "Exploitation requires knowledge of a valid public survey hash."

A public survey hash is the code at the end of a public survey link. Its whole job is to be handed out. Study teams paste it into recruitment emails, flyers, QR codes, patient portals, social posts and course pages. Securifera's write-up says it plainly: public survey links "are built for unauthenticated distribution and are easily recoverable using google dorks, urlscan, or wayback machine." The hash is not a password and it was never meant to be one.

So the honest reading of CISA's precondition is: if your REDCap has ever run a public survey that anyone outside your building could see, assume an attacker can get a hash.

REDCap is where universities, academic medical centers and clinical research groups keep study data. The REDCap project's own site, read today, lists 8,539 institutions in 167 countries, 2.9 million projects and 4.7 million users. Those are the consortium's counts of members and accounts, not a count of internet-exposed or vulnerable servers. We do not have that number and we found none published in the sources we read.

What is confirmed

These statements come from the primary records: the CISA KEV entry (catalog version 2026.11.18), the CVE record published by Securifera, the NVD entry, and Securifera's technical write-up.

The flaw. NVD's description: "An unauthenticated Remote Code Execution vulnerability was found in the survey passthrough routing and Data Import processing logic." An attacker manipulates HTTP requests "to access an unintended controller route from a public survey context" and supplies "a crafted file-path/stream parameter during import handling." The weaknesses listed are CWE-73 (external control of file name or path) and CWE-94 (code injection).

The score. Securifera rates it CVSS 3.1 9.8, Critical: network attack, low complexity, no privileges, no user interaction, high impact on confidentiality, integrity and availability. CISA's own assessment in the CVE record says exploitation "active," automatable "yes," technical impact "total."

Who is affected. NVD: "This vulnerability exists in REDCap 13.3.0 and higher." The CVE record lists 13.3.0 as affected and 16.0.49, 17.3.10 and 17.4.4 as unaffected. It does not classify versions older than 13.3.0.

How it works, in one paragraph. Securifera found five separate faults that chain together. The public survey page has a passthrough feature that is supposed to reach only one PDF route; a double URL-encoded parameter makes two checks disagree, and the request reaches REDCap's general dispatcher instead. REDCap then calculates the visitor's rights before it knows the visitor is an anonymous survey respondent, and the code treats that undefined state as a super user, so the anonymous request is granted the Data Import tool. Data Import trusts a client-supplied filename, which can be a PHP stream rather than a file, and a byte-order-mark cleanup step writes data back to that same location. Securifera used that write to replace a predictable compiled template file in REDCap's temp cache, which REDCap then runs as PHP when a public page renders it. Their summary: "The extra decode opened the route. Data Import supplied the write. BladeOne supplied the execution sink."

The finder. The CVE record credits Ryan Wincey of Securifera. Securifera's write-up, published Sunday, September 27, 2026, says the research used an AI coding agent pointed at a copy of the REDCap source that a consortium member had published by accident. REDCap is not openly distributed, so the attacker pool is not limited to consortium members: leaked copies exist.

The timeline. The CVE was reserved September 13 and published Sunday, September 20, 2026. CISA added it to KEV on November 18, 59 days later.

What is not known

Who is exploiting it, and against whom. CISA's entry does not say, and we found no public report of in-the-wild exploitation from Vanderbilt, Securifera, a victim or another vendor. CISA is the only source we found for the exploitation claim.

Whether ransomware crews are using it. KEV lists known ransomware campaign use as "Unknown."

Vanderbilt's own advisory. The KEV entry points to the REDCap Community site for vendor guidance. That site is login-gated: the project's page says it "is only open to registered REDCap administrators from current partner sites." We could not read it. The fixed versions above come from the CVE record and NVD, which are public; the Community site may hold release notes or guidance we have not seen. If you are a REDCap administrator, read it.

Whether the public exploit code works. Our exploit tracker recorded five GitHub repositories named for this CVE, the first two on September 21, the day after the CVE went public, and one from Securifera's own account on October 1. We did not open or run any of them. Repositories that appear within a day of a CVE are sometimes fakes or malware lures aimed at defenders; do not run one on a machine you care about.

Indicators. None published in the sources we read: no IP addresses, domains or file hashes appear in the KEV entry, NVD, the CVE record or Securifera's write-up. We added nothing to our feed for this flaw.

What to do

  1. Upgrade. REDCap 16.0.49 LTS, 17.3.10 LTS or 17.4.4 Standard Release, or later on the same track. If you are on a version older than these on any track, you are in the affected range the CVE record describes.
  2. If you cannot upgrade today, turn off public surveys or put them behind your institution's login, and block survey-context requests to the dispatcher, Data Import and Parcel routes at your reverse proxy or WAF. Then upgrade.
  3. Assume your survey hashes are known. Search urlscan and the Wayback Machine for your own REDCap hostname and see how many of your public survey links are already sitting there. That is what an attacker sees.
  4. Look for signs it already happened. See the self-check below. CISA's forensic triage flag on this entry is a hint that it expects some agencies to find something.
  5. If you find a compromise, treat the study data as exposed. REDCap often holds identifiable participant and patient data, which turns a server incident into a reporting question for your IRB, privacy office and, where it applies, HIPAA breach rules.

Am I affected? A concrete self-check

Your version. REDCap serves each version from a folder named after it, so the version appears in the page URLs and on disk as redcap_v followed by the number, for example redcap_v17.3.9. Log in, look at the address bar on any project page, or look at the folder names in your REDCap web root. The Control Center home page also shows the installed version. Below 16.0.49 on the 16.0 LTS track, below 17.3.10 on the 17.3 LTS track, or below 17.4.4 on the standard track: affected.

Your access logs. These search terms come from the defender guidance in Securifera's write-up. Grep your web server access logs since September 20 for:

  • __passthru with an encoded value, especially %25 (a double-encoded character). A normal passthrough request has a plain value.
  • DataImportController or ParcelController in a request that also carries a survey hash parameter (s=). Those routes have no business being reached from a public survey.
  • Any request where a fname value contains php:// or the word filter, if your logs capture POST bodies or your WAF does.

Your cache folder. List the files under REDCap's temp folder in the cache subfolder (the files end in .bladec) and sort by modification time. A compiled template that changed outside a REDCap upgrade, or one containing PHP that is not template output, deserves a closer look.

Your processes. Look for shells, curl, wget or interpreters started by the web server or PHP-FPM account. REDCap does not normally spawn them.

We have not published a checker for this flaw. The steps above are the check.

How we know, and what our edge saw

The KEV dates come from CISA's KEV JSON feed, catalog version 2026.11.18, downloaded today. The weekdays and the 59-day gap come from the system calendar, not from press reports. The CVE details come from the NVD API and the CVE.org record, and the technical chain from Securifera's own advisory list and write-up. We wrote this from those primaries; we did not rely on press coverage.

Our edge saw nothing, and that tells you little. We scanned our edge telemetry since August 1 for the REDCap survey, passthrough, Data Import and versioned-folder paths, plus the path fragments our tracker pulled from the public repositories. Zero matches in 565,766 edge block records, zero in 163,957 edge canary records, zero in 5,717 origin honeypot records. But we do not run REDCap and we have no REDCap decoy, so attackers looking for REDCap servers have no reason to knock on our door. Our edge could not have seen this campaign. That is a blind spot, not a clean bill of health.

We have no detection lead here and claim none. The public repositories our tracker logged are third-party code and a clock, not our detection. CISA's listing is the news, and the fix has been out since September.

Sources

CISA Known Exploited Vulnerabilities catalog, entry CVE-2026-90817, date added November 18, 2026, due November 21, 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-90817

NVD, CVE-2026-90817 (published September 20, 2026; CVSS 3.1 9.8 from Securifera; CISA SSVC exploitation "active"): https://nvd.nist.gov/vuln/detail/CVE-2026-90817

CVE.org record published by Securifera as the assigning authority (affected and fixed versions, solution text, credit): https://www.cve.org/CVERecord?id=CVE-2026-90817

Securifera advisories list: https://www.securifera.com/advisories/

Securifera, "Survey Says…. Remote Code Execution: CVE-2026-90817," September 27, 2026: https://www.securifera.com/blog/2026/09/27/survey-says-remote-code-execution-cve-2026-90817/

REDCap Community Site (vendor guidance linked from KEV; login required, we could not read it): https://projectredcap.org/resources/community

Read next: our post on the Puppet Enterprise flaw CISA added yesterday, another case where a credential the attacker is "supposed" to lack turns out to be easy to get: https://www.dugganusa.com/post/puppet-enterprise-cve-2026-115731-exploited-kev


Every indicator in this post is in the feed. Free.

1.58M+ IOCs, STIX 2.1 / TAXII, 88% novel vs ThreatFox, exploited-CVE leads ahead of CISA. No credit card — a free API key in 30 seconds, and you can audit every claim above against the live endpoints.

Get your free key → analytics.dugganusa.com/stix/register